swap space problem

[leyearn]

system physical memory was 2G
swap space is 3G
now we added the memory to 5G
but now when I execute #vmstat
and in /var/adm/syslog.dated/*/kern.log

Jan 20 13:42:56 yj8201 vmunix: swap space below 10 percent free
Jan 20 13:43:37 yj8201 last message repeated 3 times
Jan 20 13:45:39 yj8201 last message repeated 9 times
Jan 20 13:55:29 yj8201 last message repeated 36 times

but when i execute #swapon -s

yj8201#swapon -s
swap space below 10 percent free
Swap partition /dev/rz8b (default swap):
Allocated space: 262144 pages (2048MB)
In-use space: 1 pages ( 0%)
Free space: 262143 pages ( 99%)

Swap partition /dev/rz10b:
Allocated space: 125000 pages (976MB)
In-use space: 1 pages ( 0%)
Free space: 124999 pages ( 99%)


Total swap allocation:
Allocated space: 387144 pages (3024MB)
Reserved space: 365180 pages ( 94%)
In-use space: 2 pages ( 0%)
Available space: 21964 pages ( 5%)

[Ralf Puchner]

if adding memory, adjust swapspace. Have a look into the manual how memory is calculated and used, change memory allocation algorithm!

Formular: swapspace = 2.5 times physical memory

Have a look to the "reserved memory" information!

[Venkatesh BL]

The total swap space allocated appear to be 3 GB. You've mentioned that you have a total of 5 GB as swap. What happened to the remaining 2 GB?

[Johan Brusche]

Seems you have eager(=immediate mode) swapping enabled.
(See "ls -l /sbin/swapdefault" ) and maybe
you (or the DBA)told Oracle instances to use a larger SGA.(why else would you have added memory to the system ;-)

With the additional 2GB, you probably can go for using "deferred" swap-mode. To do this, rename the /sbin/swapdefault link to something like /sbin/swapdefault.off , and reboot.

Read the swapon manpage for additional info.

__ Johan.

[Victor Semaska_3]

A warning about using deferred mode. This from section 3.6.3 of the System Administration Guide:

The operating system terminates a process if it attempts to write a modified virtual page to swap space that is depleted.

In other words, if you use deferred mode and the swap space isn't large enough, the OS will start deleting processes.

Vic

[Han Pilmeyer]

I wouldn't worry too much about that statement in 3.6.3 in the documentation. When your total vitual address exceeds is too large for your swap space (and memory in case of lazy swap) processes will start to die anyway. It's just a matter of which process dies first.

For the case in this topic, process start to fail when total virtual memory in the system exceeds the total swap space (somewhere around 3 GB). If deferred (lazy) swap was enabled, processes would start to die when total virtual memory exceeds total swap + total memory (3 GB + 5 GB = 8 GB).

If use of internal memory was over committed by 3 GB on the swap space, then the system would probably be so slow that you would welcome processes being stopped.

[Victor Semaska_3]

Just letting him/her know the possibility exists. We've had it happen. Some utility went 'nuts' and used vast amounts of virtual memory. Ran out of swap space and that's when system processes started to disappear. Had to reboot the server to get it back to normal. Fortunately it wasn't a production server running mission critical apps.

Vic

[Ralf Puchner]

the documentation contains all the suggested improvements and explanations...